Implementation of R-GCNs for Relational Link Prediction
Clone or download
Latest commit 722eaa8 Sep 21, 2018
Type Name Latest commit message Commit time
Failed to load latest commit information.
.idea update readme Feb 19, 2018
code Update Sep 21, 2018
data small fixes May 22, 2018
settings small fixes May 22, 2018
.gitignore small fixes May 22, 2018
LICENSE Added license Feb 19, 2018 Update Feb 19, 2018 small fixes May 22, 2018 Added slurm scripts Mar 17, 2017

Graph Convolutional Networks for Relational Link Prediction

This repository contains a TensorFlow implementation of Relational Graph Convolutional Networks (R-GCN), as well as experiments on relational link prediction. The description of the model and the results can be found in out paper:

Modeling Relational Data with Graph Convolutional Networks. Michael Schlichtkrull, Thomas N. Kipf, Peter Bloem, Rianne van den Berg, Ivan Titov, Max Welling (ArXiv 2017)


  • TensorFlow (>1.0)

Running demo

We provide a bash script to run a demo of our code. In the folder settings, a collection of configuration files can be found. The block diagonal model used in our paper is represented through the configuration file settings/gcn_block.exp. To run a given experiment, execute our bash script as follows:

bash \[configuration\]

We advise running the model on a modern GPU, as training can take up to several hours.


Please cite our paper if you use this code in your own work:

  title={Modeling Relational Data with Graph Convolutional Networks},
  author={Schlichtkrull, Michael and Kipf, Thomas N and Bloem, Peter and Berg, Rianne van den and Titov, Ivan and Welling, Max},
  journal={arXiv preprint arXiv:1703.06103},